Interactions between cougars, humans, and other animals present challenges in regions with high cougar activity. In the proposed study, data from the Washington Department of Fish and Wildlife (WDFW) and a novel cougar activity dataset will be used to investigate how clustering algorithms can be used to uncover patterns in cougar behavior and whether classifiers can predict high-impact cougar events. This project aims to enhance existing knowledge of cougar aggression behavior with the possibility of offering new insights, ultimately leading to improved public safety and wildlife conservation efforts.
The first dataset utilized in this research is the publically available Washington Department of Fish and Wildlife (WDFW) dataset. This dataset contains reports of cougar safety confrontations. Only reports that included confirmed sightings of cougars were included in this study. Additionally, extensive data cleaning was done due to inconsistent labeling and missing data. Figure 1 below gives a basic summary of the locations of the reports from 2016 to 2025.
Figure 1: The map above illustrates the count of confirmed cougar safety confrontations reported to the WDFW from 2016 to 2025, with darker colors indicating more reports.
In Figure 1, Stevens, Spokane, and Chelan county stand out as having the highest number of cougar safety confrontation reports. However, it is important to note that factors such as county population, topography, and land cover can affect these values.
In addition to the cleaned WDFW dataset, a second dataset was collected using private Facebook posts of trail cameras in Washington State. Unlike the safety confrontation reports of the first dataset, this dataset is largely made up of cougar trail camera sightings, providing a different perspective of overall cougar movements. Figure 2 below shows a map of the most popular GMUs for 2024 based on a sample, while Figure 3 provides a visualization of cougar Facebook post dates based on an incomplete sample of the second cougar activity dataset.
Figure 2: The map above shows the number of verified cougar sightings reported with sufficient information to a private Facebook group during 2024. Black missing sections on the map represent national parks and indian reservations where GMUs do not exist.
Figure 3: The plot above displays cougar sightings over time, separated by Game Management Unit (GMU). The size of each circle represents the number of cougars sighted on that date in that GMU.
Using both of the datasets described above, cougar activity and aggression data will be clustered to reveal patterns in cougar behavior. Additionally, earlier activity and other contextual features will be used in machine learning classifiers to predict future cougar activities.